Homemade LED light source is working

[Eric Woodbury]

Eric, nice work. Briefly checked the schematic on your blog; looks like a quite thoroughly engineered solution. I see you're using photodiodes to sense the output of the green & blue leds; is this done to compensate for aging and component drift?
Can you comment on the purpose of R6, R55, R48 and R24? Where does the flash circuit connect to in the rest of the schematic? I see you're running single supply on the opamps; did you witness any issues with a small offset when M1 is conducting, or do the MCP6002's switch off nicely with In+ held down to GND?
Also, did you consider using LED drivers such as the MP24894, and/or was there a particular reason not to do this?

Forgot this schematic was there. I should update. Many changes, but I will address your questions anyway. R6, R55, F48, R24 are 'power dropping' resistors. I use a 36V power supply. LEDs voltage drop is pretty much invariant and I didn't want all the excess voltage across the FETs. These resistors are the backup plan. The green and blue LEDs have slightly different Vforward, so the resistors can help with balance here, too. In the end, I wanted about 2V across the FETs during highest current. Turned out that the power supply has some slight adjustment in output voltage, so maybe one could get away without these resistors.

I removed the photo diode circuits in the final version. I was anxious to get it working and the optics of separating blue/green diodes with filters and such seemed fraught with problems. Further analysis with more knowledge of how much heating I have suggests that the photo diodes were unnecessary. Those circuits are now simple followers.

Yes, it is single supply. Notice that there are some offsets imposed on some opamps to prevent offsets and tolerancing from being a problem. It's like having a minus 50 mV second supply.

I ended up using LM324 equivalents. Many opamps would work. I didn't use switchers because these tend to be surface mount parts and I built my circuits by hand. SMDs are too small for me. I have a client for whom I'm designing an LED driver system. We use over a thousand LED switchers in that behemoth. Pure analog is easier to make and fix.

M1 replaced with a couple of NPNs. Vsat is low enough to extinguish the LEDs Vref below the offsets. FLASH only needs the 5V and gnd. I put it here only so I didn't need another box for another control. Also borrowed power for red LED to illuminate my f/numbers on the enlarger lens.

[koraks]

Thanks Eric; that makes good sense. After posting I indeed figured the resistors were current dropping resistors to offload dissipation from the fets.
Copy that on the virtual ground at 50mV for some of the opamps; that does the trick. I also understand the choice for an NPN instead of M1; the reason I asked is that I found that the Rdson of the 2n7000 is relatively high (around 2R as I recall) which can pose problems according to some simulations I had coincidentally been doing over the weekend on a concept somewhat similar to yours. In my case I used the 2n7000 (actually a 7002; the smd variant) to pull the reference voltage of a current limiter to ground by means of a PWM signal.

What problems did you run into with the transimpedance amplifiers for the photodiodes? That part of the schematic piqued my interest as I've been messing about with something similar lately as well. Although in my case they're not linear amplifiers, but integrating ones because I need to accurately measure a PWM light source. It also needs to interface with a microcontroller, so the output is fed into an ADC. I have a working bench prototype at the moment, but it's still a little noisy - particularly it still picks up some 50Hz hum (this is Europe).

I understand the challenge of getting a good readout for green and blue separately in your application. It would be tricky to integrate overall output of the combined green vs. blue arrays, without 'contamination' between green and blue. I thought about it for a bit, but it seems there's no easy solution. I also think it's not absolutely necessary, especially for B&W work. I did experience something that could be related to a heating issue when doing color with leds, where the output of the red channel seemed to increase after an initial heat-up period of the red leds. I worked around it by keeping the red leds off in focus/compose mode and only turn them on for the actual printing exposure. In the new version of the system I'm working on I hope to prevent this by providing better thermal management, but the problem may still rise its ugly head (although the workaround is fairly simple and seems effective). When doing B&W, I noticed no issues with non-linearity that posed a problem - turns out the contrast response of the paper is the main source of variability that easily drowns any minor issues with the light source...(i.e. different emulsion speeds for high/low contrast emulsions, resulting in a distinct sensitivity peak around the center grades, low sensitivity at low contrast and somewhat higher sensitivity at high contrast).


Btw, I understand about the choice of DIP/through-hole instead of SMD. I recently decided to include SMD as well, which led me to start making my own PCB's as well. Since it's not a commercial product, I can get away with fairly ugly contraptions. Turns out I actually quite like DIY PCB-making and soldering SMD stuff...more than a little crazy!?